That's fine for consumer stuff but people like me doing CAD and stuff still want all the performance they can get. I'm not going to have as much motivation to drop huge money on new compute servers and workstations if the performance gains aren't significant. The machines we are buying now pack two Xeon E5-2690's that run about $2k each, that's $4k straight to Intel of the ~8k price!
More performance per watt is obviously good, it means more cores per chip with the same power consumption.
Unfortunately however not every computation is threaded to take advantage of many cores. We see this in the gaming scene where a dual core i3 can beat an 8-core AMD CPU because it's single-thread performance is superior. »www.tomshardware.com/reviews/far ··· 9-7.html
And let's not forget that in the consumer market, gamers are the ones dropping the big cash for high-end CPUs on relatively short upgrade cycles.
I think much of this could be addressed in software. Not sure about the games but I know for a fact that calculations that could be threaded in the software I use are not. For example I want to analyze a circuit at 1000 different frequencies, and I bang my head on the desk as I watch 1 core of my 16-core process each frequency in order. That is a blatant example that could be fixed pretty easily, but there are other problems that are much harder to break up.
(One piece of good news on the gaming front is that both next-gen Xbox and Playstation are rumoured to be packing 8-core AMD CPUs. This may give game developers the kick they need to develop increasingly multi-threaded games, and x86/64 compatible to boot.)
In summary performance, and thread performance, are still important to many consumers and businesses, not to mention the consumers and businesses which are buying the highest quantities of Intel's most expensive processors.